Update README.md

README.md

@@ -9,13 +9,18 @@ tags:
 - reasoning
 - llm
 - DIRA
+- qwen
+- unsloth
+- transformers
 ---
 
 # Diraya-3B-Instruct-Ar
 
 ## Model Description
 
-Diraya-3B-Instruct-Ar is an Arabic reasoning-specialized language model fine-tuned from Qwen2.5-3B. This model is part of the DIRA (Diraya Arabic Reasoning AI) collection, which focuses on enhancing the logical inference and mathematical reasoning capabilities of Arabic language models.
+**Diraya-3B-Instruct-Ar** is an `Arabic` reasoning-specialized language model fine-tuned from `Qwen2.5-3B` . 
+
+This model is part of the **DIRA (Diraya Arabic Reasoning AI)** collection, which focuses on enhancing the logical inference and mathematical reasoning capabilities of **Arabic** language models.
 
 ## Key Features
 
@@ -31,7 +36,6 @@ Diraya-3B-Instruct-Ar is an Arabic reasoning-specialized language model fine-tun
 
 **Model Type**: Instruction-tuned causal language model
 
-**Parameter Count**: 3.09B (2.77B non-embedding)
 
 **Architecture**:
 - 36 transformer layers
@@ -40,7 +44,7 @@ Diraya-3B-Instruct-Ar is an Arabic reasoning-specialized language model fine-tun
 - Context length: 32,768 tokens
 
 **Training Approach**:
-- Fine-tuned using GPRO (General Policy Reinforcement Optimization)
+- Fine-tuned using `GPRO` 
 - Training focused on structured reasoning output format using XML tags
 - Optimized for mathematical reasoning using the Arabic GSM8K dataset
 - Multiple reward functions including correctness, format adherence, and output structure
@@ -77,12 +81,16 @@ The model is designed to output structured reasoning in the following format:
 ### Example Usage
 
 ```python
-from transformers import AutoModelForCausalLM, AutoTokenizer
+from unsloth import FastLanguageModel
+
+model, tokenizer = FastLanguageModel.from_pretrained(
+    model_name = "Omartificial-Intelligence-Space/Diraya-3B-Instruct-Ar",
+    max_seq_length = max_seq_length,
+    load_in_4bit = True, # False for LoRA 16bit
+    fast_inference = True, # Enable vLLM fast inference
+    max_lora_rank = lora_rank,
+)
 
-# Load the model and tokenizer
-model_name = "Omartificial-Intelligence-Space/Diraya-3B-Instruct-Ar"
-model = AutoModelForCausalLM.from_pretrained(model_name)
-tokenizer = AutoTokenizer.from_pretrained(model_name)
 
 # System prompt to enforce XML structure
 system_prompt = """
@@ -121,8 +129,8 @@ print(response)
 
 This model was primarily fine-tuned on:
 
-- **Arabic GSM8K Dataset**: A comprehensive collection of grade school math problems translated to Arabic, requiring multi-step reasoning
-- **Format**: Training emphasized structured reasoning using XML tags to clearly separate reasoning steps from final answers
+- [**Arabic GSM8K Dataset**](https://huggingface.co/datasets/Omartificial-Intelligence-Space/Arabic-gsm8k):
+- A comprehensive collection of grade school math problems translated to Arabic, requiring multi-step reasoning
 
 ## Training and Evaluation Results
 
@@ -146,23 +154,6 @@ The model demonstrates strong performance on Arabic mathematical reasoning tasks
 - Following the required XML output format
 - Arriving at correct numerical answers for multi-step problems
 
-## Limitations
-
-- Specialized for reasoning tasks and may not perform as well on general conversational tasks
-- Performance may vary on complex mathematical problems beyond grade-school level
-- Limited to the Arabic language
-
-## Responsible Use
-
-This model is intended for educational and research purposes. While it excels at mathematical reasoning, please note:
-- It should not replace human judgment for critical decisions
-- Results should be verified when used in educational contexts
-- The model inherits limitations from its base model Qwen2.5-3B
-
-## Related Resources
-
-This model is part of the DIRA (Diraya Arabic Reasoning AI) collection:
-- [Arabic GSM8K Dataset](https://huggingface.co/datasets/Omartificial-Intelligence-Space/Arabic-gsm8k): The dataset used for training this model
 
 ## Citation
 
@@ -196,4 +187,4 @@ This model builds upon the Qwen2.5-3B model by the Qwen Team and utilizes optimi
       journal={arXiv preprint arXiv:2407.10671},
       year={2024}
 }
-```
+```